Optimization of Distributed Storage System of Legal Electronic Data Based on Machine Learning

With the advent of the big data era, the volume of electronic data in the legal service system grows dramatically, while the amount of data access gradually increases, resulting in slower and slower response times for system functions. Therefore, to solve the response time problem of legal electronic data, this work proposes a performance tuning method based on distributed storage system. Firstly, the distributed storage features are screened twice to obtain the set of parameters that affect the storage performance. In the process of performing training sample generation, orthogonal experimental design method is used to select representative feature samples for experiments to obtain the experimental values of throughput and latency. Then, the training samples are preprocessed, and the parameter samples and experimental values are combined as the sample data for the prediction model. An integrated neural network model based on Kernel Density Estimation (KDE) is proposed for the problem of sample data containing unbiased noise. The integrated neural network model is used to train and learn from the sample data, and further feature selection is performed to obtain throughput and delay prediction models. Finally, according to the characteristics of the throughput and delay prediction models, multiple swarms co-evolutionary approach is used to improve the position updating method of the traditional Fruit fly optimisation algorithm. The performance of the distributed storage system is optimised using the hybrid Fruit fly optimisation algorithm, and the optimal solution as well as the corresponding optimal parameter configurations are obtained. Experiments are conducted using four typical workloads to verify the accuracy of the proposed prediction model. The results show that the proposed tuning method can effectively improve the performance of the distributed storage system for legal electronic data. © 2024, Taiwan Ubiquitous Information CO LTD. All rights reserved.